aboutsummaryrefslogtreecommitdiffstats
path: root/src/mesa/drivers/dri/i965/brw_wm.c
blob: bedfec13f235eacf9d4e01ee29fa81e84ac0cbd5 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
/*
 * Copyright (C) Intel Corp.  2006.  All Rights Reserved.
 * Intel funded Tungsten Graphics to
 * develop this 3D driver.
 *
 * Permission is hereby granted, free of charge, to any person obtaining
 * a copy of this software and associated documentation files (the
 * "Software"), to deal in the Software without restriction, including
 * without limitation the rights to use, copy, modify, merge, publish,
 * distribute, sublicense, and/or sell copies of the Software, and to
 * permit persons to whom the Software is furnished to do so, subject to
 * the following conditions:
 *
 * The above copyright notice and this permission notice (including the
 * next paragraph) shall be included in all copies or substantial
 * portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
 * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
 * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
 * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
 * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 */
#include "brw_context.h"
#include "brw_wm.h"
#include "brw_state.h"
#include "main/enums.h"
#include "main/formats.h"
#include "main/fbobject.h"
#include "main/samplerobj.h"
#include "main/framebuffer.h"
#include "program/prog_parameter.h"
#include "program/program.h"
#include "intel_mipmap_tree.h"
#include "intel_image.h"
#include "compiler/brw_nir.h"
#include "brw_program.h"

#include "util/ralloc.h"

static void
assign_fs_binding_table_offsets(const struct gen_device_info *devinfo,
                                const struct gl_program *prog,
                                const struct brw_wm_prog_key *key,
                                struct brw_wm_prog_data *prog_data)
{
   uint32_t next_binding_table_offset = 0;

   /* If there are no color regions, we still perform an FB write to a null
    * renderbuffer, which we place at surface index 0.
    */
   prog_data->binding_table.render_target_start = next_binding_table_offset;
   next_binding_table_offset += MAX2(key->nr_color_regions, 1);

   next_binding_table_offset =
      brw_assign_common_binding_table_offsets(devinfo, prog, &prog_data->base,
                                              next_binding_table_offset);

   if (prog->nir->info->outputs_read && !key->coherent_fb_fetch) {
      prog_data->binding_table.render_target_read_start =
         next_binding_table_offset;
      next_binding_table_offset += key->nr_color_regions;
   }
}

static void
brw_wm_debug_recompile(struct brw_context *brw, struct gl_program *prog,
                       const struct brw_wm_prog_key *key)
{
   perf_debug("Recompiling fragment shader for program %d\n", prog->Id);

   bool found = false;
   const struct brw_wm_prog_key *old_key =
      brw_find_previous_compile(&brw->cache, BRW_CACHE_FS_PROG,
                                key->program_string_id);

   if (!old_key) {
      perf_debug("  Didn't find previous compile in the shader cache for debug\n");
      return;
   }

   found |= key_debug(brw, "alphatest, computed depth, depth test, or "
                      "depth write",
                      old_key->iz_lookup, key->iz_lookup);
   found |= key_debug(brw, "depth statistics",
                      old_key->stats_wm, key->stats_wm);
   found |= key_debug(brw, "flat shading",
                      old_key->flat_shade, key->flat_shade);
   found |= key_debug(brw, "per-sample interpolation",
                      old_key->persample_interp, key->persample_interp);
   found |= key_debug(brw, "number of color buffers",
                      old_key->nr_color_regions, key->nr_color_regions);
   found |= key_debug(brw, "MRT alpha test or alpha-to-coverage",
                      old_key->replicate_alpha, key->replicate_alpha);
   found |= key_debug(brw, "fragment color clamping",
                      old_key->clamp_fragment_color, key->clamp_fragment_color);
   found |= key_debug(brw, "multisampled FBO",
                      old_key->multisample_fbo, key->multisample_fbo);
   found |= key_debug(brw, "line smoothing",
                      old_key->line_aa, key->line_aa);
   found |= key_debug(brw, "input slots valid",
                      old_key->input_slots_valid, key->input_slots_valid);
   found |= key_debug(brw, "mrt alpha test function",
                      old_key->alpha_test_func, key->alpha_test_func);
   found |= key_debug(brw, "mrt alpha test reference value",
                      old_key->alpha_test_ref, key->alpha_test_ref);

   found |= brw_debug_recompile_sampler_key(brw, &old_key->tex, &key->tex);

   if (!found) {
      perf_debug("  Something else\n");
   }
}

/**
 * All Mesa program -> GPU code generation goes through this function.
 * Depending on the instructions used (i.e. flow control instructions)
 * we'll use one of two code generators.
 */
static bool
brw_codegen_wm_prog(struct brw_context *brw,
                    struct brw_program *fp,
                    struct brw_wm_prog_key *key,
                    struct brw_vue_map *vue_map)
{
   const struct gen_device_info *devinfo = &brw->screen->devinfo;
   struct gl_context *ctx = &brw->ctx;
   void *mem_ctx = ralloc_context(NULL);
   struct brw_wm_prog_data prog_data;
   const GLuint *program;
   GLuint program_size;
   bool start_busy = false;
   double start_time = 0;

   memset(&prog_data, 0, sizeof(prog_data));

   /* Use ALT floating point mode for ARB programs so that 0^0 == 1. */
   if (fp->program.is_arb_asm)
      prog_data.base.use_alt_mode = true;

   assign_fs_binding_table_offsets(devinfo, &fp->program, key, &prog_data);

   /* Allocate the references to the uniforms that will end up in the
    * prog_data associated with the compiled program, and which will be freed
    * by the state cache.
    */
   int param_count = fp->program.nir->num_uniforms / 4;
   prog_data.base.nr_image_params = fp->program.info.num_images;
   /* The backend also sometimes adds params for texture size. */
   param_count += 2 * ctx->Const.Program[MESA_SHADER_FRAGMENT].MaxTextureImageUnits;
   prog_data.base.param =
      rzalloc_array(NULL, const gl_constant_value *, param_count);
   prog_data.base.pull_param =
      rzalloc_array(NULL, const gl_constant_value *, param_count);
   prog_data.base.image_param =
      rzalloc_array(NULL, struct brw_image_param,
                    prog_data.base.nr_image_params);
   prog_data.base.nr_params = param_count;

   if (!fp->program.is_arb_asm) {
      brw_nir_setup_glsl_uniforms(fp->program.nir, &fp->program,
                                  &prog_data.base, true);
   } else {
      brw_nir_setup_arb_uniforms(fp->program.nir, &fp->program,
                                 &prog_data.base);

      if (unlikely(INTEL_DEBUG & DEBUG_WM))
         brw_dump_arb_asm("fragment", &fp->program);
   }

   if (unlikely(brw->perf_debug)) {
      start_busy = (brw->batch.last_bo &&
                    drm_bacon_bo_busy(brw->batch.last_bo));
      start_time = get_time();
   }

   int st_index8 = -1, st_index16 = -1;
   if (INTEL_DEBUG & DEBUG_SHADER_TIME) {
      st_index8 = brw_get_shader_time_index(brw, &fp->program, ST_FS8,
                                            !fp->program.is_arb_asm);
      st_index16 = brw_get_shader_time_index(brw, &fp->program, ST_FS16,
                                             !fp->program.is_arb_asm);
   }

   char *error_str = NULL;
   program = brw_compile_fs(brw->screen->compiler, brw, mem_ctx,
                            key, &prog_data, fp->program.nir,
                            &fp->program, st_index8, st_index16,
                            true, brw->use_rep_send, vue_map,
                            &program_size, &error_str);

   if (program == NULL) {
      if (!fp->program.is_arb_asm) {
         fp->program.sh.data->LinkStatus = linking_failure;
         ralloc_strcat(&fp->program.sh.data->InfoLog, error_str);
      }

      _mesa_problem(NULL, "Failed to compile fragment shader: %s\n", error_str);

      ralloc_free(mem_ctx);
      return false;
   }

   if (unlikely(brw->perf_debug)) {
      if (fp->compiled_once)
         brw_wm_debug_recompile(brw, &fp->program, key);
      fp->compiled_once = true;

      if (start_busy && !drm_bacon_bo_busy(brw->batch.last_bo)) {
         perf_debug("FS compile took %.03f ms and stalled the GPU\n",
                    (get_time() - start_time) * 1000);
      }
   }

   brw_alloc_stage_scratch(brw, &brw->wm.base,
                           prog_data.base.total_scratch,
                           devinfo->max_wm_threads);

   if (unlikely((INTEL_DEBUG & DEBUG_WM) && fp->program.is_arb_asm))
      fprintf(stderr, "\n");

   brw_upload_cache(&brw->cache, BRW_CACHE_FS_PROG,
                    key, sizeof(struct brw_wm_prog_key),
                    program, program_size,
                    &prog_data, sizeof(prog_data),
                    &brw->wm.base.prog_offset, &brw->wm.base.prog_data);

   ralloc_free(mem_ctx);

   return true;
}

bool
brw_debug_recompile_sampler_key(struct brw_context *brw,
                                const struct brw_sampler_prog_key_data *old_key,
                                const struct brw_sampler_prog_key_data *key)
{
   bool found = false;

   for (unsigned int i = 0; i < MAX_SAMPLERS; i++) {
      found |= key_debug(brw, "EXT_texture_swizzle or DEPTH_TEXTURE_MODE",
                         old_key->swizzles[i], key->swizzles[i]);
   }
   found |= key_debug(brw, "GL_CLAMP enabled on any texture unit's 1st coordinate",
                      old_key->gl_clamp_mask[0], key->gl_clamp_mask[0]);
   found |= key_debug(brw, "GL_CLAMP enabled on any texture unit's 2nd coordinate",
                      old_key->gl_clamp_mask[1], key->gl_clamp_mask[1]);
   found |= key_debug(brw, "GL_CLAMP enabled on any texture unit's 3rd coordinate",
                      old_key->gl_clamp_mask[2], key->gl_clamp_mask[2]);
   found |= key_debug(brw, "gather channel quirk on any texture unit",
                      old_key->gather_channel_quirk_mask, key->gather_channel_quirk_mask);
   found |= key_debug(brw, "compressed multisample layout",
                      old_key->compressed_multisample_layout_mask,
                      key->compressed_multisample_layout_mask);
   found |= key_debug(brw, "16x msaa",
                      old_key->msaa_16,
                      key->msaa_16);

   found |= key_debug(brw, "y_uv image bound",
                      old_key->y_uv_image_mask,
                      key->y_uv_image_mask);
   found |= key_debug(brw, "y_u_v image bound",
                      old_key->y_u_v_image_mask,
                      key->y_u_v_image_mask);
   found |= key_debug(brw, "yx_xuxv image bound",
                      old_key->yx_xuxv_image_mask,
                      key->yx_xuxv_image_mask);

   for (unsigned int i = 0; i < MAX_SAMPLERS; i++) {
      found |= key_debug(brw, "textureGather workarounds",
                         old_key->gen6_gather_wa[i], key->gen6_gather_wa[i]);
   }

   return found;
}

static uint8_t
gen6_gather_workaround(GLenum internalformat)
{
   switch (internalformat) {
   case GL_R8I: return WA_SIGN | WA_8BIT;
   case GL_R8UI: return WA_8BIT;
   case GL_R16I: return WA_SIGN | WA_16BIT;
   case GL_R16UI: return WA_16BIT;
   default:
      /* Note that even though GL_R32I and GL_R32UI have format overrides in
       * the surface state, there is no shader w/a required.
       */
      return 0;
   }
}

void
brw_populate_sampler_prog_key_data(struct gl_context *ctx,
                                   const struct gl_program *prog,
                                   struct brw_sampler_prog_key_data *key)
{
   struct brw_context *brw = brw_context(ctx);
   GLbitfield mask = prog->SamplersUsed;

   while (mask) {
      const int s = u_bit_scan(&mask);

      key->swizzles[s] = SWIZZLE_NOOP;

      int unit_id = prog->SamplerUnits[s];
      const struct gl_texture_unit *unit = &ctx->Texture.Unit[unit_id];

      if (unit->_Current && unit->_Current->Target != GL_TEXTURE_BUFFER) {
         const struct gl_texture_object *t = unit->_Current;
         const struct gl_texture_image *img = t->Image[0][t->BaseLevel];
         struct gl_sampler_object *sampler = _mesa_get_samplerobj(ctx, unit_id);

         const bool alpha_depth = t->DepthMode == GL_ALPHA &&
            (img->_BaseFormat == GL_DEPTH_COMPONENT ||
             img->_BaseFormat == GL_DEPTH_STENCIL);

         /* Haswell handles texture swizzling as surface format overrides
          * (except for GL_ALPHA); all other platforms need MOVs in the shader.
          */
         if (alpha_depth || (brw->gen < 8 && !brw->is_haswell))
            key->swizzles[s] = brw_get_texture_swizzle(ctx, t);

         if (brw->gen < 8 &&
             sampler->MinFilter != GL_NEAREST &&
             sampler->MagFilter != GL_NEAREST) {
            if (sampler->WrapS == GL_CLAMP)
               key->gl_clamp_mask[0] |= 1 << s;
            if (sampler->WrapT == GL_CLAMP)
               key->gl_clamp_mask[1] |= 1 << s;
            if (sampler->WrapR == GL_CLAMP)
               key->gl_clamp_mask[2] |= 1 << s;
         }

         /* gather4 for RG32* is broken in multiple ways on Gen7. */
         if (brw->gen == 7 && prog->nir->info->uses_texture_gather) {
            switch (img->InternalFormat) {
            case GL_RG32I:
            case GL_RG32UI: {
               /* We have to override the format to R32G32_FLOAT_LD.
                * This means that SCS_ALPHA and SCS_ONE will return 0x3f8
                * (1.0) rather than integer 1.  This needs shader hacks.
                *
                * On Ivybridge, we whack W (alpha) to ONE in our key's
                * swizzle.  On Haswell, we look at the original texture
                * swizzle, and use XYZW with channels overridden to ONE,
                * leaving normal texture swizzling to SCS.
                */
               unsigned src_swizzle =
                  brw->is_haswell ? t->_Swizzle : key->swizzles[s];
               for (int i = 0; i < 4; i++) {
                  unsigned src_comp = GET_SWZ(src_swizzle, i);
                  if (src_comp == SWIZZLE_ONE || src_comp == SWIZZLE_W) {
                     key->swizzles[i] &= ~(0x7 << (3 * i));
                     key->swizzles[i] |= SWIZZLE_ONE << (3 * i);
                  }
               }
               /* fallthrough */
            }
            case GL_RG32F:
               /* The channel select for green doesn't work - we have to
                * request blue.  Haswell can use SCS for this, but Ivybridge
                * needs a shader workaround.
                */
               if (!brw->is_haswell)
                  key->gather_channel_quirk_mask |= 1 << s;
               break;
            }
         }

         /* Gen6's gather4 is broken for UINT/SINT; we treat them as
          * UNORM/FLOAT instead and fix it in the shader.
          */
         if (brw->gen == 6 && prog->nir->info->uses_texture_gather) {
            key->gen6_gather_wa[s] = gen6_gather_workaround(img->InternalFormat);
         }

         /* If this is a multisample sampler, and uses the CMS MSAA layout,
          * then we need to emit slightly different code to first sample the
          * MCS surface.
          */
         struct intel_texture_object *intel_tex =
            intel_texture_object((struct gl_texture_object *)t);

         /* From gen9 onwards some single sampled buffers can also be
          * compressed. These don't need ld2dms sampling along with mcs fetch.
          */
         if (brw->gen >= 7 &&
             intel_tex->mt->msaa_layout == INTEL_MSAA_LAYOUT_CMS &&
             intel_tex->mt->num_samples > 1) {
            key->compressed_multisample_layout_mask |= 1 << s;

            if (intel_tex->mt->num_samples >= 16) {
               assert(brw->gen >= 9);
               key->msaa_16 |= 1 << s;
            }
         }

         if (t->Target == GL_TEXTURE_EXTERNAL_OES && intel_tex->planar_format) {
            switch (intel_tex->planar_format->components) {
            case __DRI_IMAGE_COMPONENTS_Y_UV:
               key->y_uv_image_mask |= 1 << s;
               break;
            case __DRI_IMAGE_COMPONENTS_Y_U_V:
               key->y_u_v_image_mask |= 1 << s;
               break;
            case __DRI_IMAGE_COMPONENTS_Y_XUXV:
               key->yx_xuxv_image_mask |= 1 << s;
               break;
            default:
               break;
            }
         }

      }
   }
}

static bool
brw_wm_state_dirty(const struct brw_context *brw)
{
   return brw_state_dirty(brw,
                          _NEW_BUFFERS |
                          _NEW_COLOR |
                          _NEW_DEPTH |
                          _NEW_FRAG_CLAMP |
                          _NEW_HINT |
                          _NEW_LIGHT |
                          _NEW_LINE |
                          _NEW_MULTISAMPLE |
                          _NEW_POLYGON |
                          _NEW_STENCIL |
                          _NEW_TEXTURE,
                          BRW_NEW_FRAGMENT_PROGRAM |
                          BRW_NEW_REDUCED_PRIMITIVE |
                          BRW_NEW_STATS_WM |
                          BRW_NEW_VUE_MAP_GEOM_OUT);
}

void
brw_wm_populate_key(struct brw_context *brw, struct brw_wm_prog_key *key)
{
   struct gl_context *ctx = &brw->ctx;
   /* BRW_NEW_FRAGMENT_PROGRAM */
   const struct brw_program *fp = brw_program_const(brw->fragment_program);
   const struct gl_program *prog = (struct gl_program *) brw->fragment_program;
   GLuint lookup = 0;
   GLuint line_aa;

   memset(key, 0, sizeof(*key));

   /* Build the index for table lookup
    */
   if (brw->gen < 6) {
      /* _NEW_COLOR */
      if (prog->info.fs.uses_discard || ctx->Color.AlphaEnabled) {
         lookup |= BRW_WM_IZ_PS_KILL_ALPHATEST_BIT;
      }

      if (prog->info.outputs_written & BITFIELD64_BIT(FRAG_RESULT_DEPTH)) {
         lookup |= BRW_WM_IZ_PS_COMPUTES_DEPTH_BIT;
      }

      /* _NEW_DEPTH */
      if (ctx->Depth.Test)
         lookup |= BRW_WM_IZ_DEPTH_TEST_ENABLE_BIT;

      if (brw_depth_writes_enabled(brw))
         lookup |= BRW_WM_IZ_DEPTH_WRITE_ENABLE_BIT;

      /* _NEW_STENCIL | _NEW_BUFFERS */
      if (ctx->Stencil._Enabled) {
         lookup |= BRW_WM_IZ_STENCIL_TEST_ENABLE_BIT;

         if (ctx->Stencil.WriteMask[0] ||
             ctx->Stencil.WriteMask[ctx->Stencil._BackFace])
            lookup |= BRW_WM_IZ_STENCIL_WRITE_ENABLE_BIT;
      }
      key->iz_lookup = lookup;
   }

   line_aa = BRW_WM_AA_NEVER;

   /* _NEW_LINE, _NEW_POLYGON, BRW_NEW_REDUCED_PRIMITIVE */
   if (ctx->Line.SmoothFlag) {
      if (brw->reduced_primitive == GL_LINES) {
         line_aa = BRW_WM_AA_ALWAYS;
      }
      else if (brw->reduced_primitive == GL_TRIANGLES) {
         if (ctx->Polygon.FrontMode == GL_LINE) {
            line_aa = BRW_WM_AA_SOMETIMES;

            if (ctx->Polygon.BackMode == GL_LINE ||
                (ctx->Polygon.CullFlag &&
                 ctx->Polygon.CullFaceMode == GL_BACK))
               line_aa = BRW_WM_AA_ALWAYS;
         }
         else if (ctx->Polygon.BackMode == GL_LINE) {
            line_aa = BRW_WM_AA_SOMETIMES;

            if ((ctx->Polygon.CullFlag &&
                 ctx->Polygon.CullFaceMode == GL_FRONT))
               line_aa = BRW_WM_AA_ALWAYS;
         }
      }
   }

   key->line_aa = line_aa;

   /* _NEW_HINT */
   key->high_quality_derivatives =
      ctx->Hint.FragmentShaderDerivative == GL_NICEST;

   if (brw->gen < 6)
      key->stats_wm = brw->stats_wm;

   /* _NEW_LIGHT */
   key->flat_shade = (ctx->Light.ShadeModel == GL_FLAT);

   /* _NEW_FRAG_CLAMP | _NEW_BUFFERS */
   key->clamp_fragment_color = ctx->Color._ClampFragmentColor;

   /* _NEW_TEXTURE */
   brw_populate_sampler_prog_key_data(ctx, prog, &key->tex);

   /* _NEW_BUFFERS */
   key->nr_color_regions = ctx->DrawBuffer->_NumColorDrawBuffers;

   /* _NEW_COLOR */
   key->force_dual_color_blend = brw->dual_color_blend_by_location &&
      (ctx->Color.BlendEnabled & 1) && ctx->Color.Blend[0]._UsesDualSrc;

   /* _NEW_MULTISAMPLE, _NEW_COLOR, _NEW_BUFFERS */
   key->replicate_alpha = ctx->DrawBuffer->_NumColorDrawBuffers > 1 &&
      (_mesa_is_alpha_test_enabled(ctx) ||
       _mesa_is_alpha_to_coverage_enabled(ctx));

   /* _NEW_BUFFERS _NEW_MULTISAMPLE */
   /* Ignore sample qualifier while computing this flag. */
   if (ctx->Multisample.Enabled) {
      key->persample_interp =
         ctx->Multisample.SampleShading &&
         (ctx->Multisample.MinSampleShadingValue *
          _mesa_geometric_samples(ctx->DrawBuffer) > 1);

      key->multisample_fbo = _mesa_geometric_samples(ctx->DrawBuffer) > 1;
   }

   /* BRW_NEW_VUE_MAP_GEOM_OUT */
   if (brw->gen < 6 || _mesa_bitcount_64(prog->info.inputs_read &
                                         BRW_FS_VARYING_INPUT_MASK) > 16) {
      key->input_slots_valid = brw->vue_map_geom_out.slots_valid;
   }

   /* _NEW_COLOR | _NEW_BUFFERS */
   /* Pre-gen6, the hardware alpha test always used each render
    * target's alpha to do alpha test, as opposed to render target 0's alpha
    * like GL requires.  Fix that by building the alpha test into the
    * shader, and we'll skip enabling the fixed function alpha test.
    */
   if (brw->gen < 6 && ctx->DrawBuffer->_NumColorDrawBuffers > 1 &&
       ctx->Color.AlphaEnabled) {
      key->alpha_test_func = ctx->Color.AlphaFunc;
      key->alpha_test_ref = ctx->Color.AlphaRef;
   }

   /* The unique fragment program ID */
   key->program_string_id = fp->id;

   /* Whether reads from the framebuffer should behave coherently. */
   key->coherent_fb_fetch = ctx->Extensions.MESA_shader_framebuffer_fetch;
}

void
brw_upload_wm_prog(struct brw_context *brw)
{
   struct brw_wm_prog_key key;
   struct brw_program *fp = (struct brw_program *) brw->fragment_program;

   if (!brw_wm_state_dirty(brw))
      return;

   brw_wm_populate_key(brw, &key);

   if (!brw_search_cache(&brw->cache, BRW_CACHE_FS_PROG,
                         &key, sizeof(key),
                         &brw->wm.base.prog_offset,
                         &brw->wm.base.prog_data)) {
      bool success = brw_codegen_wm_prog(brw, fp, &key,
                                         &brw->vue_map_geom_out);
      (void) success;
      assert(success);
   }
}

bool
brw_fs_precompile(struct gl_context *ctx, struct gl_program *prog)
{
   struct brw_context *brw = brw_context(ctx);
   struct brw_wm_prog_key key;

   struct brw_program *bfp = brw_program(prog);

   memset(&key, 0, sizeof(key));

   uint64_t outputs_written = prog->info.outputs_written;

   if (brw->gen < 6) {
      if (prog->info.fs.uses_discard)
         key.iz_lookup |= BRW_WM_IZ_PS_KILL_ALPHATEST_BIT;

      if (outputs_written & BITFIELD64_BIT(FRAG_RESULT_DEPTH))
         key.iz_lookup |= BRW_WM_IZ_PS_COMPUTES_DEPTH_BIT;

      /* Just assume depth testing. */
      key.iz_lookup |= BRW_WM_IZ_DEPTH_TEST_ENABLE_BIT;
      key.iz_lookup |= BRW_WM_IZ_DEPTH_WRITE_ENABLE_BIT;
   }

   if (brw->gen < 6 || _mesa_bitcount_64(prog->info.inputs_read &
                                         BRW_FS_VARYING_INPUT_MASK) > 16) {
      key.input_slots_valid = prog->info.inputs_read | VARYING_BIT_POS;
   }

   brw_setup_tex_for_precompile(brw, &key.tex, prog);

   key.nr_color_regions = _mesa_bitcount_64(outputs_written &
         ~(BITFIELD64_BIT(FRAG_RESULT_DEPTH) |
           BITFIELD64_BIT(FRAG_RESULT_STENCIL) |
           BITFIELD64_BIT(FRAG_RESULT_SAMPLE_MASK)));

   key.program_string_id = bfp->id;

   /* Whether reads from the framebuffer should behave coherently. */
   key.coherent_fb_fetch = ctx->Extensions.MESA_shader_framebuffer_fetch;

   uint32_t old_prog_offset = brw->wm.base.prog_offset;
   struct brw_stage_prog_data *old_prog_data = brw->wm.base.prog_data;

   struct brw_vue_map vue_map;
   if (brw->gen < 6) {
      brw_compute_vue_map(&brw->screen->devinfo, &vue_map,
                          prog->info.inputs_read | VARYING_BIT_POS,
                          false);
   }

   bool success = brw_codegen_wm_prog(brw, bfp, &key, &vue_map);

   brw->wm.base.prog_offset = old_prog_offset;
   brw->wm.base.prog_data = old_prog_data;

   return success;
}